7870 ghz vs. 7950 vs. 670

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

technoholic

Distinguished
Feb 27, 2008
800
0
19,160
69
I'm leaning to believe that this frame latency stuff is being used by NV fanboys since they can find not much argument to use against AMD's GCN cards (especially at the value point). I have been making research about this mentioned issue and i see that techreport article is being widely used by fans a lot. Let's skip the "cherry picked game benches" in that article, cause tech-report somehow managed to show 7950 SO MUCH below the 660Ti. In reality, both cards perform on par especially with latest drivers from both sides. If you care to make a research in google, you'll see that most reviews in fact shows 7950 a little ahead of 660Ti at the end. (Not talking about those that use AMD powered games only)

Now, when i look at the posted videos i see NO difference between the two records. Is it only me to admit this? I personally think that this issue is being exaggerated much cause i don't believe most anyone can sense this latency, or "stutters". If there's any, GTX 660Ti ALSO shows it (watch the video before responding please). If anything, it can as well be a bug from drivers. OR maybe this is something really related to the hardware. So, it is 7950's raw horse power against 660Ti's so-called consistency. It is funny cause Fermi cards also showed some frame latency issues BUT the counter argument was always "ok but fermi produces high FPS/is fast" :)


Am i defending AMD? Yes and no. With the current pricing, i can absolutely recommend Nvidia and AMD cards, as long as they produce good enough gaming for an acceptable price.
 
Technoholic, take a look at my post about 'future-proofing' and longevity for an illustration of the importance of this. People have been talking for a while about the flaws of frames/second benchmarking, but I wasn't sure what they were talking about, so opened this thread:

http://www.tomshardware.co.uk/forum/380802-15-frames-measure-performance

Check out Scott's response (the big one) - really helpful, detailed response that explained a lot. Turns out, people have been discussing the flaws of frames/second benchmarking for over a decade since nVidia TNT2 days!

When people ask about the framerates they can get from a card, they're not really wanting to know the number of frames their GPU can render per second. They're wanting to know how smooth/jerky gameplay will be at their chosen settings. And frame latency is a much, much better measure of that smoothness.

I can understand why you'd want to cling onto frames/second - I was reluctant to accept that all those hours spent analysing benchmark charts were wasted. I was reluctant to disregard that knowledge. But personally, if I find that my knowledge is flawed, I'd rather embrace the truth and start looking at testing done with a methodology more relevant to real-world performance, so that moving forward I can give the best advice and make the best purchase decisions.

And you know me Technoholic - we've PMed about the more technical aspects of this stuff, so you know I'm not a 'fanboy' and back my recommendations up by fact. So there's at least one person who isn't a fanboy saying that there is a very real argument here for frame latency benchmarking :)

EDIT: Also, I wasn't aware of the claims that Fermi couldn't deliver consistent framerates - don't suppose you have a link for that? No worries if not.
 

regina_49

Honorable
Jun 26, 2012
437
0
10,810
10
I can agree this forum is filled with AMD and NV fans.(I think more so amd fans are really out of hand).It's just the petty, ridiculous arguments i cannot stand both offer great options.
 


Agreed - there's far too much fighting without any real solid arguments or proof. That's not what you're seeing here though. I don't believe Technoholic is a fanboy, and I'm pretty sure I'm not either :) You're looking at intelligent, reasoned and articulate arguments here, that are backed up by facts and references.

'Both sides are great' is a popular response to these discussions, but the 'everything is great' approach doesn't really help the OP (and other readers) answer the question 'which option is best for me?'. By the 'everything is great' logic, you may as well just pick up whatever if you like the price. It's important to thoroughly examine every aspect so that you're making the most informed purchase / recommendation possible.
 

burntpizza

Honorable
Jun 23, 2012
194
0
10,710
9


Except their earlier articles completely contradicts that later article, latency was the same from 660 ti to 7950. Moreover, their results contradticts about every other tech site on the planet. I wouldn't trust them on that as they haven't done enough investigation to warrant a conclusion. It could be a dud gfx card or that the present version of catalyst is being wierd.
 


I wasn't aware they got different results with the same methodology. The results are repeated and confirmed when they re-ran the numbers for Windows 7 though? You know the reason it doesn't line up with other tech sites is that they're not testing the same thing.

Certainly for frames/second, 7950 meets or beats the GTX660 Ti, but the argument is that frames/second is a flawed measure of performance. And that argument looks pretty sound to me? I'd be very interested if there is a flaw in their methodology, but I know other tech sites are planning to adopt this same testing method, so it must be worth doing?

EDIT: Also, look at this: http://www.tomshardware.co.uk/radeon-hd-7990-devil13-7970-x2,review-32563-8.html - THG got the same results with other Radeon/GeForce models. I know these are dual-GPU, but they show the same pattern of nVidia delivering far more consistent performance, even on their dual-GPU card. Across 3000 frames, even a dual-GPU GeForce gets smaller lag spikes than a single-GPU Radeon!
 

technoholic

Distinguished
Feb 27, 2008
800
0
19,160
69
Sam i trust your judgement, my words were against those who keep using this current issue as a final and the deciding argument without the ultimate understanding of what they are talking about. You know what you talk about and thanks for sharing your educating information with me, giving your precious time explaining me about all that tech stuff :) The situation here, on the other hand seems kind of complicated to me (i don't know if i express my thoughts with the right word). Are we jumping to conclusions very quickly?

Now check here:
http://techreport.com/review/21516/inside-the-second-a-new-look-at-game-benchmarking/10

And please read the final page of the review.
- The writer suggests that the high frame latency may be the result of game engine, and can show itself in different architectures
- GTX 570 there shows high latencies compared to GTX 580 (which is kind of strange because they are in the same family and they have similar specs), even higher than some SLI/X-fire setups. Dual cards normally should produce more "stutter" yes?
-And if there is only one source (tech-report) doing these tests, how can anyone make the decision? Which games are they testing, with what settings, with what driver version, who is doing these tests, are they biased/unbiased....etc?

Most people don't get what i mean. FPS may not be the ultimate factor to explain how a card performs. But frame latency tests are also not always direct and sometimes confusing. What would the final determining factor be in measuring a card's performance? Can this be a hardware or software issue? People are paying real money to buy GPUs so we must all know what is the good and bad with what we shed our money on. Is this, for example, a software/hardware trick to skip frames and show the product superior against the rivaling one?
 

regina_49

Honorable
Jun 26, 2012
437
0
10,810
10
I understand what you mean.The fact of the matter is no one can answer the question.I agree with what someone earlier said. We cannot see into the future, so who knows what future proofing is ? I honestly do not believe in it.
 


Haha yeah that was me :) My point being that the rate of advance can't be predicted. But we know for a fact that there will be an advance. We have 30 years of PC gaming history that's telling us games will become more and more demanding year by year.

It's the reason I'd never buy the bare minimum I need for today's games. Do that and you end up buying a new card every 18 months. I always buy a bit of excess, so that the hardware lasts me and I save money in the long run. Future-proofing is tricky for the reasons mentioned, but we can't ignore the fact that games will become more demanding. Just look at Crysis 3 coming up and GTA5!

EDIT: To reflect Regina's edit.
 

regina_49

Honorable
Jun 26, 2012
437
0
10,810
10
I agree with you.I've noticed many people recommend a minimum of 2GB or even 3GB of VRAM at this modest resolution and I'm of the impression that these individuals are somewhat correct. I'd think 3gb of VRAM would be more future proof ?
 


THG has a similar article, and with similar results, though they compared SLI vs Crossfire.

Also, that article has one major flaw in my opinion; it focuses too much on latency. For me, latency is huge, as it causes me simulator sickness, but for at least half the population, that isn't the case. Those who do, will have less nausea.

What you should focus on is the frame spiking. AMD frames are very inconsistent on how fast they render. You can see this by the huge bars on these charts:


You'll find the article from THG to focus on what I think is the most important, but they do it for Crossfire and SLI, but not single cards: http://www.tomshardware.co.uk/radeon-hd-7990-devil13-7970-x2,review-32563-8.html

Note: While most people won't notice a big difference or any difference between the two, it is also worth noting, most people don't notice a big difference between 5-10 FPS. The result is, we want the smoothest game play, and atm, a 660ti has smoother game play than a 7950, even if the 7950 has slightly higher FPS (though in those games they were pretty similar at stock speeds).
 

Embra

Distinguished
Aug 2, 2008
2,136
0
20,160
147
All three caeds being considered will play games at 1080p quite well.

The 670 will be the strongest of the three of course. It will OC.
The 7950 is a wicked OCer. If your considering OCing, then this card may be your best value.
The 7870 will OC, but but as much as the 7950. Probably best bang for buck price/perfomance.
 
+1 to Bystander. Regina, with regards to RAM, I think 3GB is maybe excessive. It never hurts to have a bit more, but the typical argument there is that by the time games need 3GB video RAM, today's GPUs will be so outdated and obsolete that they won't get the job done anyway. I think RAM is important for future-proofing, which is why I'd never recommend 1GB cards, but I think 2GB will be ample for as long as today's GPU's can get the job done.
 

regina_49

Honorable
Jun 26, 2012
437
0
10,810
10
o.k. thanks. :) I was somewhat concerned.I have a 670 and do not plan on upgrading for at least two years.I never payed $375 for a video card before so far it's proven to be worth it. :lol:
 

Fokissed

Distinguished
Feb 10, 2010
392
0
18,810
13
Current games are just passing over 1GB VRAM at 1080p (my 5870 1GB is constantly at 999MB in modern games). However I've seen HD texture packs eat 2GB+ of VRAM (Skyrim, Oblivion, Dragon Age). While 2GB will cover you for nearly all games until your card becomes obsolete, certain mods (HD texture packs mainly) will LOVE the extra VRAM.
Also WoW (and possibly Rift, I'm not 100%) favors nVidia cards really hard. If you play a lot of WoW, then an nVidia card will be a wise choice, as AMD and nVidia cards are neck and neck in most other games (for the price, though AMD cards have really pulled ahead with the 12.11 driver).
 


Given I've had more AMD cards over the years than Nvidia cards, perhaps very close to the same, I find it hard to call me a fanboy. I also only posted an article that should be taken into consideration, as very few game sites have ever looked at this sort of information.

Lately, all we have are AMD fanboys overwhelming the site telling us how superior their cards are. There needs to be some one giving Nvidia a fair shake. And while you might not like that Nvidia delivers smoother frames, they do and people should be ware of it when making a purchase, instead of seeing the same thing the vast majority of the sites show, FPS. If you've ever read hardocp reviews, you'll notice they've noted many times that they had what appeared to be smoother game play from Nvidia systems which had lower FPS, they just never investigated why. The only 2 sites I've ever seen do such a study was THG and techreport.

Also, latency issues are not something you can see in a video, you have to experience it. Movies never bother anyone, with even 10 second latency, as it doesn't matter how long it response to your actions, but gaming when you control the screen, latency becomes a big deal, and makes me sick. They also can't show you the variances of frame times, as they themselves only show an image 24-30 times per second, not allowing you to see the small differences between frames.
 

truprecht

Distinguished
Oct 17, 2010
102
0
18,710
6
I bought a 7950 in September after spending a lot of time trying to decide between 7870, 7950, 660ti and 670. After a while I decided the 670 just cost too much, and settled on the 7950 due mostly to larger memory bandwidth (384 vs. 192), slightly better price at the time, and slightly better performance in FPS comparisons. Shortly thereafter the prices on the 7870s came down, and Tom's took the 7950 off their best value list. Then the TechReport article about latency came out, and I felt like I made the wrong choice.

At the moment, I think the 7870 is the best value for money around $240, and the 660ti is the best performance under $350. I console myself with the thought that the 3GB RAM and 384-bit bandwidth gives the 7950 more "potential" that AMD could unlock with a future driver and/or firmware update.

Does any of this diminish my enjoyment when I play Skyrim on max settings with 10 graphics-enhancing mods? No. I have noticed that Skyrim can feel a little jumpy/twitchy and that Virtu MVP really does seem to smooth it out.

My advice is to pick one of the cards, enjoy it, and stop reading GPU reviews until your card simply will not play some new game that you absolutely have to play. Any of these cards will likely last you 3+ years if "future you" can tolerate medium settings on Crysis 5, Metro 2055 or Elder Scrolls 7.
 

technoholic

Distinguished
Feb 27, 2008
800
0
19,160
69
THG's test was done between dual card configs, no single cards there. There are certain solutions providing the fix for dual card issues. Nvidia's solution is at hardware, AMD cards get it by a 3rd party free software. Let's not confuse this issue with dual card issues. This seems to me something different. What i am investigating is about single cards and the amount of data related to this frame latency issue is still not enough to prove anything, in my eyes. Just not enough to draw a certain conclusion. Because if it was so, should we call the awesome GTX 570 a "bad" product? Was single GTX 570 (which was praised so much in benchmarks) worse even than a dual gpu set-up? Let's not forget the fact that they do these "tests" using FRAPS, which is not the best piece of software to do benchmarks. I am going to approach this issue with doubts, until wider/deeper analysis made. At last, what we consider the most important is if the piece of hardware we pay for will provide smooth experience or not. And this frame latency stuff is not so clear to the eye (happens in milliseconds). Does that mean we have only numbers and graphs to compare? Isn't this kind of vague to you too?
 

technoholic

Distinguished
Feb 27, 2008
800
0
19,160
69
Here, you will understand what i mean better:

http://techreport.com/review/22922/nvidia-geforce-gtx-670-graphics-card/5

They compare Nvidia Kepler to AMD GCN in BF3. According to this test, if you pay attention, GCN cards produce more FPS and have less frame latency in BF3. Isn't BF3 a Nvidia sponsored game?

Now should i copy those graphs and paste here and yell all around, telling Nvidia Kepler completely and entirely sucks in BF3? Hell no i can't. Should i advice people to buy AMD cards instead of Nvidia depending on these graphs? Nope, not wise. I doubt and keep questioning the validity of these tests really
 

Isaiah4110

Distinguished
Jan 12, 2012
603
0
19,010
17

First, I want to say that I think truprecht's advice is superb. You will probably be happy with any of these cards, even if you let cost be the deciding factor. If you choose one of the less expensive options then there is also always the option of dropping a second card in at a later date when the prices have come down more (assuming your motherboard can handle a Crossfire setup).

Second, my personal believe is that you will probably be happiest saving a little money and going with either the 7870 or 7950.

Third, I would recommend taking a look at these models:

XFX Double D FX-787A-CDBC Radeon HD 7870 GHz Edition Black Edition for $230
XFX Double D FX-795A-TDFC Radeon HD 7950 for $290
XFX Double D Black Edition FX-795A-TDBC Radeon HD 7950 for $295

I really like XFX cards. They seem to make very good quality products and you can't really argue with the lifetime warranty. I built a PC for a friend with an XFX Radeon 6850 (Double D) and was very surprised at how quiet it ran. Now, I didn't stress test it, so I was just listening to the idle state fan speed, but it was nearly silent. The cost difference isn't that large ($5-20) and for me it is usually still worth it knowing the company stands by their product longer than 1-3 years. Working in IT, I have seen capacitors pop on video cards, and I know I'm not the only one who has seen dust clog a GPU fan (though I haven't experienced either with an XFX card yet) and just knowing that I won't have to drop another $150-300 to replace a bad card is nice.

Anyway, that's my two cents. Hope it helps.
 

bigcyco1

Champion
Moderator
Wow!Already this thread is got out of hand.I can assure you all bystander is no fanboy.I have seen him recommend amd cards he also is not making blanket statements hes providing proof. I also agree. Lately, all we have are AMD fanboys overwhelming the site telling us how superior their cards are. There needs to be some one giving Nvidia a fair shake. @OP,before this thread becomes a "amd vs nvidia" fight,let me tell you something-
both the 7950 and GTX 670 are powerful enough to max out those games at your native resolution with silky smooth frame rates.
My advice is to pick one of the cards, stop stressing on it, enjoy it!

 


Thanks for posting the article :) BF3 is indeed an exception. I read the rest of that article though and in the other games, it's the opposite story:

http://techreport.com/review/22922/nvidia-geforce-gtx-670-graphics-card/3

http://techreport.com/review/22922/nvidia-geforce-gtx-670-graphics-card/6

But you're completely right - a lot more testing needs to be done. Fortunately, that's on the way. By the time Radeon 8000 and GTX700 series launch, we should be seeing frame latency benchmarking on those on at least these two sites, and hopefully more, so we can draw conclusions from a wider range of data.

By the way, thanks for not posting graphs :) It's much better when people post links to the article rather than 1000+ pixel high posts chock full of images!

As for RadeonPro, Mautari did some really impressive work there! Makes you wonder why AMD have been unable to do this for their own cards (they seriously need to hire this guy!).

There are some drawbacks though compared to the nVidia solution. Most notably the fact that you need to benchmark each game individually and then calibrate RadeonPro for that game according to your benchmark results. Tom's Hardware said in the review that it's a complicated and time-consuming procedure. Do it wrong (even if you're off by just 10fps) and you get some massive lag spikes (right up to 100 millisecond frames, which equates to a drop to 10fps).

With nVidia on the other hand, you don't need to do a thing - the hardware is already designed to deliver consistently smooth performance out of the box. So a Radeon with RadeonPro is definitely an improvement over normal Radeon performance, but it's still not as good as normal nVidia performance, which doesn't require you to recalibrate it for each game.

I'll leave it at that though since some readers feel we're discussing this in too much detail. Honestly I'd prefer for the OP to make that call rather than others not involved in the purchase, but I think there's probably ample detail here (and links to further detail) for the OP to make their decision.
 

Tom Burnqest

Honorable
Dec 10, 2012
254
0
10,790
2

Vsync is your friend.
 

ASK THE COMMUNITY

TRENDING THREADS